Exploring the impact of ChatGPT on Wikipedia engagement

Abstract

Background: Wikipedia is one of the most widely used websites globally, providing a major information and learning resource. Prior research suggests that its use is dominated by shallow information gathering, such as fact-checking and answering questions. The public release of ChatGPT on the 30th November 2022 introduced a new, widely adopted question-answering system, raising concerns that it could divert users away from existing platforms such as Wikipedia. Purpose: This study investigates whether the release of ChatGPT has altered patterns of informating mining and contribution on Wikipedia. Research Design: We conduct a longitudinal analysis of Wikipedia engagement before and after the launch of ChatGPT. Using pairwise-comparisons and a panel regression, we compare pre- and post-ChatGPT trends and assess the volatility of results. Study Sample: The study examines activity across 12 language editions of Wikipedia for a 36 month period from the 1st of January 2021 to the 1st of January 2024. Data Analysis: We measured page views, visitor numbers, edit counts, and editor activity. Statistical analyses included pairwise pre/post comparisons using the Wilcoxon Rank-Sum test, panel regression models, and Levene’s test on absolute variance to assess volatility. Results: We find no evidence of an overall decline in Wikipedia engagement across the four metrics studied. Instead, page views and visitor numbers increased in the period following ChatGPT’s launch. However, growth was lower in languages where ChatGPT was available compared to languages where it was not. Conclusions: Our results suggest that while Wikipedia engagement has not decreased, ChatGPT may be limiting growth in certain language editions. These findings contribute to understanding how generative AI tools affect information-seeking and contribution behaviours, with implications for collective intelligence systems and the wider Web ecosystem.

Ähnliche Arbeiten